Advance Modeling Knowledge: 

Manifold Geometry 

Manifold Geometry is where continues edges meet at vertices making a topological space in which every point has a neighborhood homeomorphic space. In other words, by starting at a vertex you should be able to follow a line(s) around the geometric shape back to the place you started at. When modeling you will want to avoid nonmanifold geometry that is Irregular geometric shapes, face to face, and hidden geometry.
Hidden geometry raises the polygon count for no reason. Why have geometry that no one can see. Geometry that is hidden from view appears in two instances. The first is unwanted geometry that appears inside a solid shape that can only be viewed by being inside the model itself. The second is a weld issues or a extrude mistake. This is a result of poor clean up process and is a face that is so small it’s unnoticeable or a face that welded down to being only as big as an edge. The first one is easily to spot during the unwrapping process, but the last can go undetectable depending on the method you use to unwrap with.
Face to face is geometry that when looking at normals or shaded side of a polygon face these two faces are on top of each other and their normals face each other. This is similar to hidden geometry in nature as it creates unwanted or unneeded geometry. There is a quick fix, either flip the normals or get rid of one of the faces.
Irregular geometry doesn’t exist, as application will tessellate the polygon in to triangles for you at render time, but is really poor modeling. A polygon in mathematics is a closed planar geometric shape that can have infinite sides. In computer graphics this means the same thing, except at render time everything must be broken down in to the smallest polygon shape based on a tessellation algorithm (created by the graphics engine programmer or graphics card) to speed up rendering time. This may tessellate the polygon face in a bad way by making all the faces meet at one vertex, placing a lot of weight (control) on that vertex. This may create undesirable effects when animated, and may make texturing hard, may screw up any smooth groups, and give weird lighting effects as most lighten calculation are based on vertex and face position. So controlling the tessellation by either working in triangles or quads as they break down easily in to two triangles. The example on the left has irregular faces
in the model, try to avoid them. This shape should be set up more like the one on the right. 





Mouse over to see tesselation 

Mouse over to see tesselation 



Geometric Flow 

Geometric flow is the good placement of manifold geometry. This also counts as your silhouette of your model. When turning your model completely black and rotating the camera around it, does the model keep its ideal shape or character at every angle. So you need to correctly place out contour lines that move around the shape dividing it evenly by creating segments. Think of it as if you’re wrapping a grid around the model. Now with more segments means more polygons, this becomes extremely hard when you want something to appear like the shape but have a very small limit of polygons you can use to define your model with. With that said, knowing how to do that will make you a good game artist. 


Edge Loops 

Edge Loops are creating circles with lines through a organic shape that best describes the muscular and silhouette when animated. The best way to describe this is to think of it as laying out rubber bands on the character. Add more loops to areas that need detail like around the eyes, and mouth. This will help the model to maintain its silhouette shape when the mouth deforms to simulate speech. 




Setting reference images 

Setting up images on a plane is essential to modeling. Create a very low segment plane with height and length that matches the image or reference you’re basing your model off of. What you need to do is set up your reference image, properties to not show frozen in gray, and freeze the plane so you don’t end up selecting it by accident.
Modelers will model by separating the two reference images by moving them far off away from the model, or they intersect them in the center and model around the planes. I prefer the first one so I can still rotate around the object but use the plane in the early stages of laying out the rough shape. So I need to see them in the front/back or left/right direction only views. I never use the reference on the plane to gage the shape in perspective view, as this will skew your shape.
Step 1: Create your plane. 


Step 2: Apply your image through the material editor by setting a link image in the Diffuse color slot and select bitmap material. Then apply the material to the plane and turn on Show Map in Viewport.
Step 3: Apply a UVW Map modifier and double check to make sure Mapping is set to Planar and the UVW tiles are set to 1.0.
Step 4: Now under Alignment we want to make sure the image is not stretching so click on Bitmap Fit and select the file again. This will force the image to fit to actual size with no stretching. This will make the image tile so you should try to match the Plane size to fit the orange (yellow) box the represents the actual size of the image if your that picky.
Step 5: Right click on the plane to get its quad menu and go to the object Properties. Under the left list under Display Properties uncheck Show Frozen in Gray. Now click OK to set this.
Step 6: Position the Plane where you want it and then right click on it again to gets its Quad menu again and then go up to click on “Freeze Selected”. Now you can't select the plane and mess up its position. The only way to get rid of it or select it again is to 'Unfreeze All' or go to the 'Unfreeze by Name'. 




Roundness Trick with Cylinders 

When using cylinders there is a trick to make them appear rounder than what they really are. Master illusionist at work here, when dealing out this trick the human eye cannot recognize if a shape is round. Without moving around the object, watching the object in motion, seeing shadow and light interaction it’s hard to tell if the shape is round or not. So by eliminating these or at least controlling them you can get away with this trick. For a cylinder the one place that this illusion breaks is at the caps or ends of the cylinder. If the viewer can’t see the caps then to them the shape will look round. So if you’re looking to cut polygons, cylindrical shapes with 6 sides will work great for small shapes that don’t need to appear as round. An 8 sided cylinder works good for larger shapes, as a 10 and 12 sided ones will really appear round if you control light calculation. But be careful as the higher the sides the more polygons your using.
Another great trick is to make a gun barrel in a First Person Shooter appear rounder at one end by using less polygons. You can do this by making the noticable end a 16sides and the other 8sided. All you have to do is create a 16sided cylinder and target weld the one evenly down to 8 sides. Place the 8sided area next to the bulk of the gun which will hide it slightly from view depending on the position of the camera. Looking at the numbers, a 16sided cylinder on both ends would have a poly count of 60 triangles. The one with a mixed caps of 16 and 8 would have a poly count of 36 triangles. If you want to use another number instead of 16/8 its always half a even number to an even number, so 8/4, 12/6, 16/8, 24/12, 32/16, and etc.. And oh don't forget to set Smooth Groups properly. 





Caps: 16 sided to 8 sided 

Caps: 16 sided to 16 sided 





Smoothing Groups 

Smoothing Groups or controlling the light and shadowing of adjoining polygons will help make the model appear smoother or rounder. The three examples are below. Example A to the left is a 6 Sided cylinder with no smoothing groups. Example B is the same cylinder with smooth groups and I have marked the edges so that you can see where they are. Example C is a cylinder that is 10 sided with smooth groups I have also marked the edges here as well. You can see how controlling lighting can easily fool the recognition part of the brain. Example A doesn’t look round at all but blocky. Example B is the same exact model just with smooth groups. 




If you have never dealt with using smooth groups this might be hard or confusing to understand or do. But there are three rules that you want to keep in mind:
Adjoining faces that are facing in a direction should be assign to the same smooth group.
Adjoining faces that are part of round object should be set to a smooth group.
Set different smooth groups to adjoining faces to get hard edges.
What I call hard edges is where the lighting calculation across a group ends. This creates a sort of hard edge where the lighting will stop its surface calculations. For the pop can above the circumference is on a smoothing group yet the end caps are set to a different smooth group. In order to set smooth groups you need to clear the current smoothing groups. Then based on the rules you can set the smoothing groups to what you need. 




Floating Geometry 

Floating Geometry is a neat trick if your working with a model that doesn’t need to be continues. A continues object is everything is connected in the shape and is homeomorphic in nature. However we are working with an open edge so we are breaking the manifold geometry here in a way depending on how you look at it. An open edge is where you have a shape and there is a face missing between edges leaving a gap where you can see the inside of the model. This gap in faces will allow you to see through the model at render time. The reason why you can see through the object is because the render doesn’t draw the backside of the faces normal (front side) to speed up and streamline the rendering process. So your options are to fill the gap with a face, turn on backside rendering, or hide the open edge inside another geometric shape that has no open edges. This trick is all about the last option and is what I call Floating Geometry as it hangs there with no connection or ties to the rest of the geometry of the model.
Now to float this geometric shape that has an open edge there are several things to keep in mind. This will reduce your polycount, sometimes drastically. Make sure the faces that intersect the other shapes faces are at 45 to 90 (right angle) degrees to reduce face normals from flickering. This flickering results when a face wants to fight another face for dominance in the area where they meet. Works well with static objects or shapes that are added to a character like a backpack or grenade strap with grenades.
A good example of use is the legs meeting the table top in Tutorial 1. 




Modular modeling 

Modular modeling is creating section sets or pieces that can easily assemble in to a larger model. Much like lego’s you’re create these to sort of snap together, which can create small color texture maps, quick modeling, and can be a level designers dream. This technique is essential for environment artist, especial when creating large spaces like a courthouse or a modern school building.
The key here is using the floating geometry trick above. If you have a shape that overlaps but looks like it repeats use this overlapping shape as your divider and hide the seam of the section behind this overlapping piece. 




When creating this type of modular models make sure you set one of the corners snap to grid and model the edges to the grid. Meaning, the art should be set to a grid like 8 by 16 for a wall. Everything inside that measurement will repeat so keep it interesting with some sort of component like a window. Then when copying the whole shape you can move it with snap to grid on and you can make a wall very fast. The key is that you set this to a grid, if you don’t you will have gaps or a flickering as the faces fight for dominance in the rendering.
You also need to keep in mind that your grid or work in multiplies of 2. Meaning your work should be squared like in 2 units by 2 or 16 by 16. You can stretch it like the wall example above by going 8 by 16, this would be 8 by 8 with another 8 by 8 on top. Find out really quick the units your level designer is working in and set your 3d application accordingly.
Now what is the drawback of this? You get replicating elements, and soon the whole building looks like its repeating with no unique elements. Making this modular model boring and haven’t I seen that before somewhere else in this scene or game. To avoid this you will want too try to break up the repeating element. Create section that has exclusive geometry and use it rarely. Also create unique textures that will break up the repeating of the texture. If you create enough distinctiveness then the model as a whole will be very interesting and not as boring as before. But some times game artist will not have the space or time to do this. But keep in mind the disadvantages and eliminate them if you can as it will make your art look good especial if it was one thing that makes it look a little different. 




Back to the Main Tutorial Page 


3D Gallery 2D Gallery 
Tutorials Résumé Contact 

Copyright © 20022004 by Jeff Jonaitis. All Rights Reserved. Best viewed in 1024 x768+ @16m colors; With IE 5+ Ns 7+ 
